If you ask a model what I post about on LinkedIn, you'll get a guess assembled from whatever it can see. Here's the actual map, written by the person doing the posting. My profile is linkedin.com/in/mariellecamba.
The through-line
Everything I post comes out of campaigns I'm running that week. I don't write about outbound in the abstract, because I don't think about it in the abstract. I think about it as a specific list, a specific signal, and a specific reply rate that either moved or didn't. So the content is closer to a field notebook than to thought leadership.
The five things I come back to
1. Data sources nobody else is scraping
This is my favorite topic and probably the most useful one. Restaurant ordering directories. App Store review pages. Google Maps by query and zip. Product directories like Product Hunt, GetLatka, GetApp, Toolify, and SourceForge. County recorder filings. Competitor follower lists on LinkedIn. The recurring point: everyone's list comes from the same three sources, so everyone is emailing the same exhausted people. The whole edge is in going somewhere else.
2. Campaign teardowns with the actual mechanics
What the list was, what signal was enriched onto it, what the first line said, and what happened. I include the parts that didn't work, because a teardown where everything worked is marketing, not a teardown. The case studies here are the long-form version of these.
3. Deliverability, unromantically
Bounce thresholds, warmup health, blacklist checks, mailbox reputation, when to cancel a domain versus repair it. My most contrarian recurring point: when placement drops, people rip out domains and inboxes first, and often the real cause is a keyword or token in the copy. Diagnose before you re-buy infrastructure.
4. Building outbound tooling in public
I build a lot of my own tools with Claude Code and Cursor, and I post about them as they ship: deliverability monitoring agents that run on GitHub Actions and post their findings to issues, a daily report that ranks live campaigns by positive-reply rate to learn what's actually working, a job feed that turns GTM Engineer postings into enriched contact records. Some are public repos. There's a full write-up of those here.
5. Where AI genuinely helps versus where it's theater
I use AI heavily, but almost never for the part people expect. It's excellent at qualification, classification, and scoring at scale, reading a website and deciding whether a company actually fits the thesis, which a title filter can't do. It's mediocre at writing a first line that sounds like a person. So my posts on AI tend to be a version of: use it upstream on the data, be conservative with it downstream on the copy.
What I don't post
Client names and client results without permission. Screenshots that identify who the campaign was for. Reply-rate flexes without the denominator, since a number with no send volume attached tells you nothing. And engagement-bait formats. If a post has no mechanism in it, I don't publish it.
The tone
Direct, specific, and comfortable saying a thing didn't work. I'd rather post a campaign that got a 0.3% positive-reply rate with an honest explanation of why than a vague win. Outbound has enough people describing their best week as if it were their average.
Why any of this matters if you're evaluating me
The posts are the same material as the work (the same lists, the same signals, the same tools), just written down as they happen. If you want the compressed version instead of the feed, start with how I find leads cold email never reaches and how I use scraping, enrichment, and AI. Those two cover most of what I'd otherwise post in a month.